[ Main page of Report | Contents of Report ]
This portion of the study explored whether administrative data could be used to better understand the use of subsidies for purchase of services and to describe the disruption, dissolution, and displacement of adoptions. The highly confidential nature of adoption data posed a major challenge to this effort, because states and agencies are often unable or unwilling to share data about adopted children of families. After extensive negotiation, we were able to obtain relevant data from two states: California and North Carolina. The analyses described in this section demonstrate what could be done in other states with similar data and suggest how modifications to administrative data systems could enhance our understanding of adoptions.
Our understanding of the relationship between adoption subsidies and other post-adoption services is limited. Administrative data and surveys indicate that adoption subsidies are commonly used. As noted earlier, data from the Adoption and Foster Care Analysis and Reporting System (AFCARS) indicate that 88 percent of children adopted in 2000 were receiving subsidies (DHHS, 2001c). Preliminary AFCARS data from 2001 suggest that the number of children receiving subsidies is rising in tandem with the number of adoptions (Penelope Maza, personal communication, August 26, 2002). Many families that could qualify for subsidies, however, do not receive them (Sedlak and Broadhurst, 1993).
Little is known about pathways on and off subsidies or the reasons for, or timing of, changes in subsidy levels. Given the many children now receiving subsidies, there is a need to examine these transitions. A key issue is the transition from a deferred (or very low) subsidy to a higher subsidy, suggesting that the family has developed the need for additional services.
| States and localities vary in their subsidy policies and provisions, and in the organization of administrative data. |
States and localities are likely to vary in the assumptions that underlie the design of their subsidy programs (Bower and Laws, 2002). Some consider that subsidies should be set at a rate sufficient to provide general support for needed services. Others set subsidy amounts at a level that can only support the basic care for a child, unless there are time-limited requests for subsidy funds to address specific problems. States also vary in terms of Medicaid access for state eligible children, payments for special services, augmented rates for particularly challenging children, and payment for respite or residential care.
There is no consistency in the organization and maintenance of adoption subsidy data. Depending on the system, data may be maintained at the county level or state level; data may be integrated with the financial system used to make foster care payments or maintained in a stand-alone system; and reasons for subsidy changes may be documented well or not at all.
Administrative data do not expressly address disruption, dissolution, or displacement of adoption. Most studies of these events have relied on case record reviews and interviews — labor intensive, costly approaches that are difficult to replicate for comparison over time. An exception is the Illinois study in which Goerge, Howard, and Yu (1996) were able to match children entering foster care to children who had previously exited foster care to adoption. They identified both previously adopted children experiencing a dissolution (about 4 percent) and those placed for adoption but reentering foster care without ever having completed the adoption (about 14 percent). This effort provided a prototype for our work with North Carolina data.
[Go to Contents]
Our analysis used North Carolina data from three sources: (1) summary information on each child receiving adoption assistance, (2) vendor payments made in the name of adopted children, and (3) records of adoption subsidy checks. Because payment data were not available prior to January 1, 1990, the study population was restricted to the 8,647 children adopted after that date. Foster care placement records were then used to identify children who either had records of adoption assistance payments or who were identified as having been adopted prior to the foster care placement. These matches were complicated by the use of different ID numbers at the time of initial out-of-home placement, at adoption, and at the time of any subsequent out-of-home placement.
Children in the study population were approximately evenly divided by gender; more than half were members of minority groups. Only 12 percent of the children were older than 11 at the time of the final decree of adoption. The vast majority (90 percent) were currently receiving some form of adoption assistance. Virtually all of the children (99.8 percent) were identified as emotionally disturbed, which meets the adoption assistance eligibility requirement for “special needs.” Seventy percent had been adopted in the past five years.
| North Carolina data on adoption subsidies and foster care were merged to identify adoption dissolutions. |
North Carolina does not identify adopted children within its foster care files; however, by combining adoption subsidy and foster care placement data, we can establish a cohort of adopted children to be followed. Data on termination of parental rights (TPR) was not available in the records used for analysis.
We utilized two lines of analyses to examine adoption dissolution in North Carolina. First, we tried to track our cohort of adopted children to see if they experienced an out-of-home placement after the final decree. Second, we looked at all children who had entered out-of-home placement since July 1, 1998, to determine whether a child was previously adopted. Although neither line of analysis was entirely satisfactory, both provided information about possibilities for further research.
Cohort analysis. Three conditions were used to define dissolution: (1) date of entry into out-of-home placement occurred at least 90 days after final adoption decree date, (2) adoption assistance was no longer being received after this placement, and (3) if permanency was achieved at end of this placement, it was achieved with someone other than primary caregiver at time of placement. Of the 8,647 children in the adoption assistance data file, only 70 of these met the dissolution criteria. Using Cox Proportional Hazards Models, we estimated the risk of adoption dissolution, by age at adoption, race, gender, and year of adoption. Older children (current age) are significantly more likely to experience dissolution than younger children. Black children are twice as likely as white children to return to placement after an adoption, and about 50 percent of these dissolutions occur within three years of adoption.
| Dissolution is most likely among older children and black children — usually within 3 years of adoption. |
The finding of a less than 1 percent dissolution rate in North Carolina must be viewed cautiously. Three interpretations are possible. First, these analyses exclude families not receiving cash assistance payments, who may represent less stable adoptive relationships. This seems unlikely, based on conversations with state officials who believe that most adopted children in the state receive cash assistance payments and comparisons between the number of children adopted in North Carolina over the past several years and the number of children receiving cash assistance payments. Second, it is possible that these data and our linking algorithms do not validly link adoption assistance records to children who reentered placement under a different ID number, either the foster care number or a newly assigned number. A third possibility is that these data actually represent events in North Carolina; given the state’s relatively low rate of reentry to foster care, a low rate of adoption dissolution may also be plausible.
This line of analyses did not produce the certain results that we expected. However, if new ID numbers were systematically and consistently assigned to all children in the state who were adopted, analyses of this type could produce results that would be useful in understanding the course of an adoption that ultimately fails.
Entry into foster care. The North Carolina longitudinal placement data files provided the source of data for the second line of adoption dissolution analyses. It uses a data element added in July 1997 as part of the AFCARS enhancement that recorded whether a child who was entering out-of-home placement had been previously adopted. Of the children entering placement between July 1997 and December 2001, 318 had been adopted previously. Over half were teenagers; 58 percent were white; and 51 percent were female. Compared to the characteristics of all children who initially entered placement during the last 10 years, legally adopted children entering placement were more likely to be white (56 percent versus 47 percent) and teenagers (66 percent versus 26 percent). About one-third of the children were reunified with their primary caretaker or exited placement to a nonremoval parent, a guardian, or a court-appointed caretaker; 17 percent left for unknown or miscellaneous other reasons; 16 percent were adopted; 10 percent were emancipated; leaving slightly more than one-fourth still in placement in April 2002. Although these analyses do not provide sufficient data to calculate a dissolution rate, they suggest a higher rate of dissolution than seen in the cohort analysis. The analyses provide some insight into the number of adoption dissolutions that occur per year and the characteristics of children who are reentering placement following an adoption.
Using the foster care placement files, we next examined the question of how many children experienced an adoption disruption, that is, had placements coded as an adoptive home but ultimately were not adopted. Among the 54,747 children entering care for the first time between July 1, 1989 and June 30, 2001, 463 had a first placement recorded as an adoptive home. A full 77 percent of these children subsequently exited placement to adoption; 5 percent remained in care. A larger group of children (2,657) entered foster care for reasons other than adoption but were subsequently placed in an adoptive home. The majority of these children (59 percent) exited placement to adoption; 10 percent remained in care. The remaining children (18 percent of initial placements and 31 percent of subsequent placements in adoptive homes) may have experienced disruptions or had changes in their adoption plans for other reasons, including reunification, emancipation, running away, or a conversion to a guardianship. At this time, we cannot determine the ultimate case status of these children who had an adoption plan, but these data will be available in future years.
Although these analyses offer some insights, note that the majority of children (65 percent) who achieve permanency through adoption are never placed in an identified “adoptive home.” These are most likely foster children who are adopted by foster parents without ever having been identified in the data system as changing status from foster to adoptive homes. Considering all adopted children, then, the data do not support an effort to precisely estimate adoption disruption rates in North Carolina. They do indicate, however, that this will become more possible in the future.
| Nearly all adopted children receive subsidies, with subsidy amounts generally stable over time. |
These analyses use data that record payments to adopted children (subsidies) or for services received by adopted children (vendor payments). Because adopted children in North Carolina receive a new client ID number after the adoption decree is final, these analyses are limited to post-finalization assistance, beginning in 1990. Almost all (94 percent) children with adoption assistance received cash payments, and close to two-thirds (61 percent) also received additional assistance in the form of payments to vendors for therapeutic or medical services or nonrecurring costs of adoption. Half of the children started receiving cash payments almost immediately after the final decree. Within 6 months of the decree, 96 percent had received their first cash assistance check. The average cash payment amount during this time period was $346 per month received for an average of 42 months. However, because most of these cases are still open, these averages may change over time since there are some increases in payments as children age. Very young children received average cash assistance payments equal to $315; the average payment for children between 6 and 12 years old was $364; for children older than 12, the average payment increased to $409.
Slightly over half (51 percent) of children had no change in their subsidy amounts over the course of their assistance period. For the remaining children the increases were not substantial. The average number of days between the first cash payment and the initial increase was almost 2 years (22 months); however, this varied by age and race of adopted child. Older children were less likely to receive subsidy increases, although this was in part because older children actually had less time in which they were eligible to receive assistance. Using survival analysis to control for this effect, we found the probability of having received an increase ranged from 20 percent of children during the first year of assistance, to 50 percent by 2.5 years. Children under 5 years of age were the most likely to have subsidy increases and to incur them more quickly.
Because many factors are related to the length of time before an increase occurs, survival analysis was used to analyze the likelihood that a subsidy increase will occur, while controlling for characteristics of adopted children and length of eligibility time. Race and age at initial payment are significantly related to the likelihood of a subsidy increase. Even though the model controls for the number of months of assistance, children who begin receiving adoption assistance before age five are much more likely to receive increased subsidy payments than older children. Other minority children are less likely to receive an increased subsidy than either white or black children.
Analysis of vendor payments indicated that half of the children with a vendor payment had the first payment within two months of the adoption decree, and three-quarters had first payment within six months of the decree. The average number of vendor payments per child was four, with amounts ranging up to $2,000. The analysis of these payments is complicated by the fact that children could receive these payments before and after the final decree, and so payments for one child could be recorded under different ID numbers. Thus, it is likely that these numbers actually underestimate the amount of vendor payments incurred by an individual child.
[Go to Contents]
Analysis of adoption subsidies drew on two sources: survey data and administrative records. Participants in the California Long-Range Adoption Study (CLAS) completed questionnaires in three waves (1990, 1992, and 1996) following adoption of children from foster care in 1988–89. Data from the CLAS study include information on a broad range of psychological, social, economic, and relational characteristics of adoptive families in California, some of which has been previously reported (Brooks, Allen, and Barth, 2002).
| California administrative data were combined with those of a survey of adopted families. |
California state data include case records completed at the time of adoption placement for children placed for adoption in 1988-89, and matching Adoption Assistance Program (AAP) records through December 2000. Data were available for 1,172 cases with AAP changes during that time, of whom 771 had available case record information. A total of 401 cases were excluded because children were over age 18, cases did not match, or cases had substantial missing data. Approximately half of children were female. Birth mother’s race was most often white not Hispanic (61 percent), followed by 23 percent of Hispanic origin and 14 percent of African-American. Most of the adopting mothers were high school graduates (29 percent) or had some college or trade school (35 percent); just over half (51 percent) of adopting mothers worked outside the home prior to the adoption.
The AAP is theoretically updated with each biannual recertification or any time when the AAP amount changes; as with most administrative databases, some information is incomplete or missing. Children with many subsidy changes or those who have been in group care may be overrepresented in the database because workers have more opportunities to update their records.
| Youth receiving subsidies were more likely than others to have behavior problems in the clinical range. |
Additional analyses on subsidy use were conducted for this report to examine whether children’s behavior is associated with early changes in AAP payments. Of the 288 adopted foster children in this sample, there were exactly equal numbers (144) of those who received and those who did not receive AAP funds within 2 years of placement in their adoptive homes. AAP receipt or nonreceipt tended to remain stable over the subsequent 6 years of data collection. Youth receiving AAP throughout the study period were much more likely to have Behavior Problem Index (BPI) scores in the clinical range than those who did not receive AAP. Among those families that initiated AAP between Waves 1 and 2, the proportion with high BPI scores was 21 percent at Wave 1 and 73 percent at Wave 2.
Although limited by the small numbers of cases, these data suggest that while some families do manage to care for children with high levels of behavior problems without subsidies, the likelihood of having a subsidy and maintaining it is greater for those families with children who score in the problem behavior range. Families are more likely to transition from no subsidy to subsidy because behavior problems increase, although the reasons that families stop their subsidy use are less clear.
AAP changes may occur following required biannual recertifications, reflecting routine age-related increases; or they may result from special requests for needed services. Nearly three-fourths (73 percent) of cases had one or two payment changes to recertify or change AAP amounts during the 11-year period covered by the AAP data. The vast majority of changes filed were a result of recertifications. Further, the greatest proportion of those who had any changes had only one change during this 10-year period. Clearly adoption subsidy payments in California are, on the whole, quite stable.
| Subsidy changes tend to occur in conjunction with required recertifications. |
Additional analyses addressed the direction and size of subsidy changes. These analyses excluded payment changes that occurred when the child aged out of the adoption subsidy program at age 18. We divided the amounts of payment changes into payment increases and payment decreases, and also looked into the total average amount of each payment change.
The first monthly payment was $404, on average. The average size of the payment changes grew from the first payment change to the fifth payment change. The average amount of each payment change was just $95 monthly. Of all payment changes, 26 percent were reductions in payments, which appear to have been made to correct increases that were too high or meant to be temporary. The average payment change increases in size as the number of payment changes grows: among all first payment changes, 68 percent were gains or losses of less than $100. This proportion dropped only slightly (to 64 percent) by the third payment change, but by the fifth payment change, only 38 percent of payment changes were of that size.
Reasons for AAP changes were recoded into four categories of rate changes: (1) basic care, (2) basic and special care, (3) special care, and (4) residential care, which account for 98 percent of all rate changes.
We examined reasons for payment change during each payment change. The percentage of AAP recipients needing special care and residential care changes consistently increased from the first payment change to the fifth payment change. These data suggest it is unusual for children to have high payment changes ($500 or more) as their first payment change. Most children entering residential care do so after several payment changes requested by families to help them provide services to their children. This makes the provision of residential care seem somewhat less costly than it would be if this was a common first payment change.
Multivariate analyses were based on the subset of 771 children for whom adoption case record data were available. This subset was similar to the larger population in the proportion receiving subsidy changes, the amount and direction of change, and reasons for change. Unlike in North Carolina, most AAP recipients experience periodic payment changes, probably coinciding with recertification.
| Family income and maternal education were associated with subsidy increases. |
Bivariate relationships between case characteristics and payment changes. Analyses of bivariate associations between changes in subsidy level and adoptive families’ demographic characteristics focused on positive amount of payment changes because the negative payment changes were often in response to the positive changes. These analyses examine payment changes as events that signal needs (of varying magnitude) within the adoptive family, rather than focusing on the amount of subsidies received over time. We compared demographic differences in smaller ($0 to $300) and larger ($301 or more) amounts of monthly subsidy increases. Children adopted by a well-educated adopting mother or in higher-income families were significantly more likely to receive large amount of subsidy increase. Associations between payment change and children’s race and age were statistically insignificant. If the association between education and income holds up in the multivariate analysis, this would suggest a need for a more equitable adoption subsidy program.
Multivariate analysis: logistic regression results. We performed logistic regression analysis in order to test associations between individual demographic characteristics — after controlling for their association with other case characteristics — and the amount of payment changes. We ran three slightly different models, each one including a somewhat different combination of variables, because all variables could not be tested simultaneously and because we wanted to see whether removing education or income — which are highly correlated — affected the results. Model 1 includes the child’s race, age, and adopting mother’s educational level; model 2 includes child’s race, age, and adoptive family’s income; and model 3 includes children’s race, age, the adopting mother’s educational level, and family income. All three logistic models appeared to be significant with acceptable, but not impressive, goodness-of-fit results. However, results should be carefully considered because pseudo R² values are very small across all models, that is, the model did not explain a sizable proportion of the difference in subsidy changes. These models do not include data on child disability, which should be strongly related to subsidy amount.
Event history analysis. These analyses examine the timing of payment changes in order to understand patterns of post-adoptive services need. Many AAP recipients experienced payment change every two years because families must recertify their AAP status every two years. Only 25 percent of AAP recipients have experienced a payment change before the required two-year subsidy change. However, people who have experienced more payment changes are likely to more quickly experience other payment changes before two years. About 41 percent of AAP recipients who have experienced a fifth payment change experienced their fifth payment change before two years from the date of fourth payment change.
The probability of payment change varies by family income and race. Confirming the logistic analysis, families with incomes between $26,443 and $36,000 are significantly more likely to experience an payment change within three years after placement. Children who are of “other” races have a greater likelihood of experiencing a payment change than do white, black, and Hispanic children.
| Transition to residential care is often preceded by multiple subsidy increases. |
Transitions to residential care. Residential treatment has particular policy relevance because the federal government will not reimburse for this, but 19 states will cover some or all of its cost. Only 34 children in this sample entered residential care during the study time frame. California does not pay for for-profit residential treatment, so some children may have entered residential treatment but not be included in these data. The small group makes it impossible to estimate medians for individual variables; however, a Cox proportional hazards model could be computed. The model shows a higher likelihood of payment changes associated with residential placement for children adopted when older than three years. The number of payment changes was also significantly related to a payment change for residential treatment. Most children who entered residential treatment had three or more prior payment changes. Families with income between $36,001 and $48,761 were more likely to receive a payment change for residential treatment. Neither race nor the education of the mother was significantly related to the use of subsidies for residential treatment.
[Go to Contents]
Information relevant to understanding post-adoption dynamics, post-adoption services, and subsidy use is routinely collected and underused. Because there has been so little attention to these data, we have found substantial confusion about them. This is indicative of how foster care data were kept prior to SACWIS and other innovations in foster care data use. We believe that adoption subsidy data continue to be written over in some states, so that only the current subsidy shows — there is no history, therefore, of subsidy changes. These kinds of procedures greatly weaken our chances of showing how the pattern of subsidy changes is related to adoption outcomes. Demonstrating possible uses of subsidy data is important to motivating states to do a better job of collection, storage, retrieval, and analysis.
Taken together, the analyses in this document serve several purposes. They offer a sample of the kinds of administrative data that are available to better understand post-adoption services and supports. They offer some ideas about the kinds of analyses that can be done to bring meaning to these data. They offer some substantive findings about adoption subsidies and how they are used. Finally, they offer some ideas about modifications to administrative data systems that could improve their information yield about adoption.
Differences in data availability and structure between North Carolina and California limit our ability to assess the generalizability of our findings. Yet some clear similarities and differences have emerged. Almost all (94 percent) of the children adopted from foster care in North Carolina received cash assistance subsidy payments. The amount of the cash assistance payment remained unchanged for slightly more than half of the children (51 percent). For the rest there were gradual increases in the amount of cash payment that appear to occur as the child grows older.
| Though inconclusive, these analyses suggest possible uses of administrative data for adoption research. |
These stable subsidy amounts appear to differ from those in California, where only 17 percent of the children for whom we had data had never had a payment change. Many of these payment changes are routine subsidy increases — resulting from biannual recertification requirements — but there also appear to be fewer cases in which there are no changes. The probability of a payment change is associated with the prior number of payment changes. As prior payment changes occur, the rapidity of subsequent changes increases. Thus the number of payment changes provided could be used as a marker for outreach to families who may need additional guidance or assistance.
Relatively large subsidy increases in California are also associated with a few family characteristics — specifically, the child’s age at the time of adoption and family income. Families at middle income levels are the most likely to obtain larger subsidy increases. Also, families that have more-educated mothers obtain larger subsidies. CLAS data suggest that subsidy increases are associated with the worsening of children’s behavior; we also see that they are strongly associated with parental characteristics. The equitability of adoption subsidy adjustments needs to be better understood.
Data in North Carolina support previous findings of low dissolution rates. Although the results suggest that the risk of adoption dissolution in North Carolina is lower than that seen elsewhere, further analyses show that the risk is greater for older children and for minority children compared with infants and white children in the state. We were unable to study disruption or displacement rates in North Carolina.
In California, we could study the transition from home to residential/group treatment for the relatively small proportion of children who used this option. Event history analysis indicates that age at placement, the number of prior payment changes, and — to a lesser extent — family income are associated with state-funded residential care.
Adoption data are highly confidential and fragmented. Data about foster care histories and foster care payment amounts, adoption home studies (or their electronic summaries), adoption subsidy amounts, payments for special services (i.e., vendor payments), and disruptions, dissolutions, or displacements are often collected and stored in unrelated data systems, if at all. Record matching is often required because common identifiers do not exist.
| Confidentiality concerns, incompatible data systems, and incomplete data limit analysis. |
Data on adoption assistance in North Carolina provide a clear estimate of the payment amount and length of time that children receive cash subsidy payments. The picture of vendor payments is less clear because the overall summary data maintains year-to-date estimates rather than career estimates of payments for each child. No reasons for subsidy changes or vendor payments are included in the data that we used. Nevertheless, even with these identified data constraints, these analyses do provide an important first look at these critical issues and begin to identify ways in which administrative data files might be modified to support future analyses.
The California analyses also provide important information about data issues. First, the subsidy data are not as complete as could be hoped — some children who have subsidy changes are not included in the database, as this information does not always get sent from the counties to the state. Second, there is no field in the AAP database that indicates the starting subsidy amount — all that can be gleaned from these data are the subsidy amounts upon the first payment change. Third, these data cannot be readily linked back to the foster care data, so critical information about foster care histories is not available for explaining subsequent subsidy use.
Main Page of Report | Contents of Report
Home Pages:
Human Services Policy (HSP)
Assistant Secretary for Planning and Evaluation
(ASPE)
U.S. Department of Health and Human Services
(HHS)